By Amy Sprague
October 20, 2020
The high stakes of power estimation
ISE graduate student William Yang understands the steep consequences of misestimating daily supply and demand for power for an energy company: blackouts or blown budgets.
On the demand side, not generating enough power, even by a little, will trigger blackouts for some customers, which is a terrible scenario. But overestimating demand with a costly powering up of not-needed generators is not sustainable financially. Each day, power system operators decide how many generators to turn on for the next day such that there is sufficient power flowing to everyone while while containing costs.
On the supply side, the increasing mix of renewables, while great for reducing carbon emissions, adds to the complexity of planning. Renewables, particularly wind and solar, are notorious for their supply fluctuations. A good planning method is crucial to ensure that the varying demand each day can be met by fairly unpredictable supply, all while minimizing costs and maintaining high quality service.
The development of a power stabilizing algorithm
Yang notes that the core of this planning is to avoid the worst case scenario through a method called “robust optimization.” He explains, “We want to avoid having a blackout where we are clearly not satisfying demand, so using robust optimization is a very conservative approach to hedge against these scenarios while still staying within our budget. It builds in a conservative reserve of power to stabilize power delivery and emphasizes the risk aversion to not providing enough power.”
Right now, Yang is building an algorithm to account for fluctuations in renewable supply and testing it for reliability. A next step will be to run the algorithm on real data from power companies and compare results to their current modeling systems.
The increasing urgency of this research
ISE professor Chaoyue Zhao, Yang’s adviser, explains that with many states mandating an increasing share of renewables in their energy supply, power companies will need tools to help them plan.
She says, “Washington state, with an impressive 80 percent of energy coming from hydroelectric power, is fortunate in terms of planning. Hydro is relatively stable to predict. California, by contrast, illustrates how important this work is. California’s utilities are currently incorporating about 33 percent renewables, with goals to increase to 60 percent by 2030 and 100 percent by 2045. Solar and wind account for about 20 percent of supply today, so getting a planning model in place to incorporate more of this supply will be critical to success.”
Energy delivery isn’t the only area that it makes sense to use robust optimization. There are many essential services, particularly in the public sector, that must prioritize service delivery. In particular, emergency vaccine delivery would have many parallels, particularly essential delivery with a possibly unpredictable supply and a flexible delivery capacity that needs to be put in place in advance. Yang says, “The great thing about this research is that the theory and techniques can be applied to various domains in society to make a real impact.”